Scientists at the University of Pittsburgh are throwing a serious dose of skepticism into the personalized medicine industry. While using genetic tests to determine one’s susceptibility to a disease and/or treatment sounds like a great idea, in practice we’re a long ways away from effectively implementing the technology.
From a statement by the University to Pittsburgh:
The study focused on single nucleotide polymorphisms, or SNPs — variations in short DNA sequences that have been linked to the presence of particular diseases, and that exist in the millions in the human genome. A number of companies currently offer individualized estimates for disease risks based on genome-wide SNP genotyping. These tests typically scan 500,000 to 1 million SNPs, searching for only a handful associated with a specific disease.
Dr. Weeks and colleagues focused their study on diseases for which there are strongly associated genetic variants: age-related macular degeneration, type 2 diabetes, prostate cancer, cardiovascular disease and Crohn’s disease. They found that a strong genetic association did not guarantee they could accurately discriminate between actual disease cases and controls in both mathematical models and real-world examples.
Part of the problem may be a statistical one. To provide meaningful insights, a test for disease risk needs to accurately identify positive cases and, at the same time, provide a low false positive rate. One of the challenges with current approaches to genetic testing is that they are based on a very small number of common variants, “making it likely that you will identify people at high risk who may not be at risk at all,” said Dr. Weeks. “With such a small pool of variants, it’s difficult to develop a very meaningful test for predicting disease risk.”
In addition, he said, few health care providers have adequate genetics training to make sense of the risk calculations now commercially offered and to advise their patients accordingly.
Press release: Are we selling personalized medicine before its time?